Dongguan City is located at 22°39′ to 23°09′ North latitude and 113°31′ to 114°15′ East longitude, in the south-central part of Guangdong Province, the East Bank of the Pearl River Estuary and the Pearl River Delta downstream of the Dongjiang River.

Since 2010, water spinach in towns and districts of Qingxi, Shatian, Chashan, Daojiao, and Wanjiang and other places in Dongguan City has suffered from a new disease (vividly called “pockmark” by farmers). In this study, the five sampling method was used to investigate the incidence rate of the disease in the summer of 2019 and 2020. Briefly, five points in each field and 50 clumps of water spinach at each point were examined. A total of 2,500 clumps in 10 fields were investigated. The incidence of disease was calculated as the percentage of infected clumps in total clumps surveyed. For the analysis of microbial diversity and pathogen isolation, samples from Wanjiang District were collected.

Genomic DNA of symptomatic leaf (1 g), water (10 ml), and soils (5 g) were, respectively, extracted using ALFA-SEQ Advanced Plant DNA Kit (for the leaf) and ALFA-SEQ Advanced Soil DNA Kit (for water and soil samples) (Guangdong Magigene Biotechnology Co., Guangzhou, China) according to the manual protocols. Integrity, purity, and concentration of the DNA were examined using 1% agarose gel electrophoresis. The V3–V4 region of the 16S rDNA of the sample DNA was amplified using TaKaRa Premix Taq^® version 2.0 (TaKaRa Biotechnology Co., Dalian, China) with barcoded primers (338F 5′-ACTCCTACGGGAGGCAGCA-3′; 806R 5′-GGACTACHVGGGTWTCTAAT-3′) in conditions of pre-denaturation at 94°C for 5 min; 30 cycles of 94°C for 30 s, 52°C for 30 s, and 72°C for 30 s; extension at 72°C for 10 min on BioRad S1000 (Bio-Rad Laboratory, CA, United States). Three replicates were amplified for each sample and combined for further analysis. The amplicons were checked using 1% agarose gel electrophoresis, and fragments in size of 470 bp were purified using E.Z.N.A.^® Gel Extraction Kit (Omega, United States). A library was built according to the standard process of NEBNext^® Ultra^TM DNA Library Prep Kit for Illumina^® (New England Biolabs, United States). The products were sequenced by Illumina Nova 6000 PE250 technology at Guangdong Magigene Biotechnology Co., Guangzhou, China.

Paired-end clean reads were obtained after quality control by removing the primer sequences of the paired-end raw reads using the cutadapt software¹ and assembled using arch-fastq_mergepairs V10 (parameters include the minimum overlap length as 16 bp and the maximum allowable mismatch of overlap region of splicing sequence as 5 bp)² to obtain raw tags, which were subjected to quality tailoring (-W 4 -M 20) using fastp version 0.14.1³ to get the clean tags. The raw sequencing reads were deposited in the NCBI Sequence Read Archive database under the accession number: PRJNA739543. Operational taxonomic units (OTUs) were sorted and counted using the UPARSE software. Representative sequences of each OUT were annotated by the Ribosomal Database Project (RDP) to get the species sources. Community analysis and different abundance of OTUs were performed using STAMP 2.0.8⁴, vegan (R package).

Microbial separation from the symptomatic leaves was carried out using the method previously described (Li et al., 2020). First, diseased samples were cut into small pieces (0.5–1 cm²), and the surface was disinfected in sequence with 70% ethanol solution for 30 s, 5% sodium hypochlorite for 1 min, and sterile water washing for three times. Second, tissues were placed onto Luria–Bertani (LB, containing 10 g/L of typtone, 5 g/L of yeast extract, and 10 g/L of NaCl) agar (1.5% w/v) plates and potato dextrose agar (PDA, TOPBIO, Zhaoyuan, China) plates for incubation at 28°C for 24 h and 3 days, respectively. Colonies from the plates were, respectively, streaked onto fresh LB medium plates for purification and grown in LB liquid medium with shaking at 200 rpm overnight to preserve for further study.

Bacteria were grown in LB medium until OD₆₀₀ = 1.0, and genomic DNAs were extracted using the EasyPure Bacteria Genomic DNA Kit (TransGen Biotech, Beijing, China). The 16S rDNA gene sequences were amplified using the primers 27f and 1492r (Coenye et al., 1999) listed in Supplementary Table 1. The products were examined using 1% agarose gel electrophoresis and sent to Sangon Biotech Company in Shanghai, China, for sequencing. SeqMan V.5.00 was used to assemble sequences generated from forward and reverse primers.

Single colonies of the tested strains were grown in 10 ml of LB medium at 28°C overnight and adjusted to OD₆₀₀ = 1.5. Since the leaves of water spinach are very thin and tender, inoculation by acupuncture or injection in vitro or in vivo resulted in rapid rot (data not shown). Therefore, we have created a new inoculation method suitable for pathogenicity tests on tender leaves. Briefly, leaves with a similar size on the plants grown in a pot were chosen and wiped with 70% ethanol solution on the back. After drying, sterile toothbrushes dipped in each bacterial culture were used to brush gently on the lower side of three leaves. LB liquid medium was used as a negative control. The pots were kept in a growth chamber (Shanghai YiHeng Scientific Instruments Co., Ltd) with controlled conditions as 28 ± 2°C, 75 ± 15% relative humidity, and 12-h white light illuminance (7,350 lx). Photos were taken after 2- and 7-day post inoculation. The experiment was repeated twice.

To fulfill Koch’s postulates, the bacteria were re-isolated from the diseased leaves, and the 16S rDNA gene sequences were amplified and compared with the corresponding sequences of the inoculated ones.

For co-inoculation of TC2-1 and TC3-1, sterile toothbrushes were used to gently brush on the right half lower side of the blades to make microwounds, and 100 μl of bacterial cultures (grown in LB medium until OD₆₀₀ = 1.5) of TC2-1, TC3-1, or TC2-1 + TC3-1 mixture was evenly spread onto the whole lower side of each blade. The pots were kept in the growth chamber. Photos were taken after 1- and 3-day post inoculation. The experiment was repeated three times.

To test the pathogenicity of X. perforans on different plants, the tested strains were grown in LB medium until OD₆₀₀ = 1.0, and different plant organs were selected using different inoculation methods. For pepper (Capsicum annuum) and tomato (Lycopersicon esculentum) leaves, and mango (Mangifera indica) fruit, sterile toothbrushes dipped in TC2-1 culture were used to brush gently on the inoculated sites. For citrus (Citrus reticulata) leaves, TC2-1 was inoculated on the lower side by pressing a sterile puncher (5 mm) dipped with bacterial culture and incubated at 28°C. All trays were kept at 28°C until symptoms appeared. The same volume of LB medium was inoculated as a negative control. Each assay was repeated three times.

Bacteria were grown in LB medium until OD₆₀₀ = 1.0 (greater than 10⁹ CFU/ml) (Kido et al., 2010). Leaves of Nicotiana tabacum variant K326 (Chen et al., 2018) were inoculated on the lower side by pressing a sterilized puncher (5 mm) dipped with bacterial culture and incubated at 28°C. A positive hypersensitive response (HR) reaction was recorded when the inoculated leaf tissue collapsed or light brown necrosis occurred within 48 h after inoculation (Lana et al., 2012). Each assay was repeated three times.

Pathogenic isolates that fulfilled Koch’s postulates were further identified using the multilocus sequence analysis (MLSA). For the Xanthomonas isolates, the atpD (ATP synthase β subunit) (Brady et al., 2008), avrBs2 (Type III secretion effector AvrBs2) (Hajri et al., 2009), gyrB (DNA gyrase subunit B) (Parkinson et al., 2007), cpn60 (60-kDa chaperonin protein subunits) (Hill et al., 2004), and rpoD (RNA polymerase sigma factor) (Tian, 2018) genes were selected for PCR amplification; for the Pantoea isolates, the atpD, gyrB, infB (initiation translation factor 2), and rpoB (DNA-directed RNA polymerase subunit beta) (Brady et al., 2008) gene sequences were, respectively, amplified using the primers listed in Supplementary Table 1. Amplicons were purified with a PCR Purification Kit (TransGen Biotech, Beijing, China) and sequenced by Sangon Biotech Company in Shanghai, China. Sequences were submitted to the GenBank database with accession nos. indicated in Supplementary Table 2.

In view of the classification confusion and evolution inconsistency of different housekeeping genes in Xanthomonas genus, phylogenetic trees of isolates TC1-1 and TC2-1 were built based on each gene sequence to reveal their taxonomic status; for isolates TC3-1 and TC3-2, phylogenetic analysis was performed based on the joint sequences of atpD, gyrB, infB, and rpoB genes of the two isolates and their corresponding close-related strains obtained from NCBI database. Sequences were aligned with ClustalW, and trees were constructed using the MEGA 6.0 software by the maximum-likelihood method with 1,000 bootstrap replicates.

Genomic DNA was extracted from TC2-1 in LB medium culture using the SDS method. The harvested DNA was detected by the agarose gel electrophoresis and quantified by Qubit^® 2.0 Fluorometer (Thermo Fisher Scientific, Waltham, MA, United States). The complete genome was sequenced by Novogene (Tianjin, China) using the Nanopore PromethION platform and Illumina NovaSeq platform. The sequence was first assembled with long read data and then polished with both long and short read data by medaka v1.2.5 and unicycler v0.4.8 (Wick et al., 2017), respectively. The sequencing data and genome assembly have been deposited in the NCBI database under the accession no. PRJNA742079.

PGAP (2021-01-11.build5132) (Tatusova et al., 2016) was used to predict gene structure and gene function of the complete genome. Islandviewer4 (Bertelli et al., 2017) was used to predict genomic islands. Phaster (Arndt et al., 2016) was used to predict prophage. TXSScan was used to annotate secretion systems (Abby and Rocha, 2017). Effectors were predicted by aligning all protein sequence of TC2-1 using BLASTP with validated effectors in the T3SE (Hu et al., 2017), T4SE (Bi et al., 2013; Zou et al., 2013; An et al., 2018), and T6SE databases (Li et al., 2015; An et al., 2018). Hits with e-value less than 1e^–5 and qcov_hsp_perc more than 50 were identified as effectors.

Phylogenetic analysis was performed by GTDB-tk v1.4.0 (Chaumeil et al., 2020). In detail, TC2-1 was placed to a pre-built bacteria-wide phylogenetic tree according to the sequence identity of 120 single-copy conserved genes, and then ANI values between TC2-1 and its closely related representative strains on the tree was calculated by fastANI. The taxonomic identity of TC2-1 was determined based on both the phylogenetic and ANI analyses. A separate phylogenetic tree including TC2-1 representative Xanthomonas strains was extract from the output of GTDB-tk, and the ANI value between strains was calculated again by the fastANI. Tree files were operated by ete3 (Huerta-Cepas et al., 2016), and phylogenetic tree was visualized by iTOL (Letunic and Bork, 2019).

To identify TC2-1-specific genes, a phylogenetic tree of TC2-1 and 141 X. perforans retrieved from the NCBI RefSeq database was built by the method described previously (Pierre-Alain et al., 2018). Then TC2-1 and five representative strains were selected based on the phylogenetic tree, namely, X. perforans NI1 (GCF_003136155.1), X. perforans Xp11-2 (GCF_001009445.1), X. perforans AL66 (GCF_007714065.1), X. perforans GEV2392 (GCF_006979735.1), and X. perforans CFBP 7293 (GCF_001976075.1), which were analyzed by orthofinder v2.5.2 (Emms and Kelly, 2015). TC2-1-specific genes were then screened based on the orthofinder results after filtering the protein sequences according to the parameters (qcovhsps > 80% and identity > 65%) compared with those of X. perforans in the NCBI database.

Since 2010, bacterial leaf canker disease has attacked water spinach cultivated in many towns and districts of Dongguan City, Guangdong Province, China. Due to scattered field planting, it is difficult to calculate the total planting area of water spinach. Therefore, we investigated some cultivated fields in Daojiao Town, Gaobu Town, and Wanjiang District in the summer of 2019 and 2020. The occurrences of the disease on “Baijun 311” and “Baijun 611” cultivars are 100% in all the surveyed places, while the occurrences on “Qingtong 311” cultivar are 16.7, 86, and 100% in Daojiao, Wanjiang, and Gaobu, respectively (Supplementary Table 3).

When the disease occurs, some leaves become yellow (Figure 1A), while some remain green, with brown or chlorotic spots observed on both (Figure 1B). Small raised black spots are observed on the lower side of the leaf (Figure 1C). In the early phase of infection, water-soaked transparent spots emerge on the lower side of the leaves, protuberant, edge clear (Figure 1D). Chlorotic spots appear on the upper side of the leaves, flat, edge dim, and irregular, and the middle of the lesions gradually turns brown (Figure 1E). In the following course of disease development, more and more watery spots grow on the lower side of the leaves, gradually expand, and turn brown, with a yellow halo on the edge (Figure 1F). Chlorotic spots turn larger on the leaf surface with brown in the middle and irregular yellow halo on the edge (Figure 1G). Finally, some disease spots expand and become connected; some leaves curl up and deform; the edge of the spots protrudes into crater -like cankers on the lower side (Figure 1H), which appear as brown necrosis with a yellow halo around on the upper side (Figure 1I). During the whole period of the disease, the roots and the stems remain healthy, indicating that the pathogens preferentially colonize mesophyll tissue.

Leaves and rhizospheric soil of diseased plants and irrigation water near the planting field were collected from Dafen Zone, Wanjiang District, and the microbiomes were determined by 16S rDNA amplicon sequencing. A total of about 84,000 raw paired-end reads were generated for each sample (Supplementary Table 4). The proportions of high-quality reads were between 3.23 and 25.74% for leaf samples, between 49.38 and 55.40% for soil samples, and as high as 84.83% for the water sample (Supplementary Table 4 and Supplementary Figure 1). A combined analysis of high-quality reads from all samples identified a total of 4,903 OTUs, among which 333 were detected in the four leaf tissues, 1,100 were detected in the water sample, and 4,268 were detected in the three soil samples.

The most abundant OTUs are vastly different among samples (Supplementary Table 5). In leaf samples, Xanthomonas was the genus with the highest relative abundance in the community (ranging from 88.64 to 98.88%), and Pantoea ranked second in two of the samples. In contrast, in the water and soil samples, the proportion of Xanthomonas was between 0.06 and 0.39%, and the proportion of Pantoea was between 0 and 0.15% (Figure 2 and Supplementary Table 5).

Diseased leaf samples collected from Wanjiang District were subjected to pathogen isolation. As a result, a total of 23 isolates were obtained from the leaf tissues, including 5 Xanthomonas sp. isolates (TC1-1, TC1-2, TC2-1, TC2-3, and TCX21), 3 Pantoea sp. isolates (TC3-1, TC3-2, and TC3-5), 1 Pseudomonas aeruginosa isolate (TC11), and 3 additional Pseudomonas sp. isolates (TC1-3, TC3-3, and TC3-4), as well as 1 Bacillus aquimaris isolate (TC1-4) and 10 additional Bacillus spp. isolates (TC2-2, TC2-4, TCX22, TCX31, TCX32, TCX33, WJ2-1, WJ1, WJ2-2, and WJ3) (Supplementary Table 6). For Xanthomonas isolates, all the 16S rDNA gene sequences are completely identical to those of X. axonopodis pv. commiphoreae strain LMG26789, X. perforans strains 91-118 and LH3, and 99.93% identical to that of X. axonopodis pv. citrumelo F1. For three Pantoea isolates, the 16S rDNA gene sequences are completely identical to that of Pantoea ananatis SGAir0210, and 99.73% are identical to that of P. ananatis PP1.

To determine which isolates are pathogenic, we performed pathogenicity tests of TC1-1, TC2-1, TC1-3, TC1-4, TCX11, WJ2-1, TC3-1, TC3-2, WJ1, WJ2-2, and WJ3 on leaves of water spinach, and found that only isolates TC1-1, TC2-1, TC3-1, and TC3-2 are pathogenic (Figure 3). Two days after inoculation, dense water-soaked spots raised from the back of the leaves inoculated with Xanthomonas isolates TC1-1 and TC2-1, and the spots are chlorotic and yellow with clear edge on the leaf surface (Figure 3). On the other hand, small chlorotic spots appeared on the back of the leaves inoculated with Pantoea isolates TC3-1 and TC3-2; lesions were yellow, irregular, and indistinct with non-swollen mesophyll tissue on the leaf surface (Figure 3). At 7 days post inoculation, the protuberant spots in the TC1-1- and TC2-1-inoculated groups turned brown, and some expanded together with a yellow halo around the lesions. In general, the number of spots did not increase significantly (Figure 3). In comparison, in the TC3-1-and TC3-2-inoculated groups, the disease spots rapidly spread and connected into blocks with mesophyll tissue rotted and turned brown. The leaves were seriously yellowing on the inoculated site (Figure 3). According to the disease symptoms of the inoculated leaves, those inoculated with isolates TC1-1 and TC2-1 were similar to the typical symptoms of the bacterial canker disease. Thus, we observed Xanthomonas isolates TC1-1 and TC2-1 as the pathogens of the bacterial leaf canker of water spinach.

The simultaneous isolation of both Pantoea and Xanthomonas from diseased plants raises the possibility of synergistic or antagonistic interaction between both pathogens. Thus, co-inoculation of TC2-1 and TC3-1 were performed, and results showed that in the condition of wound pre-treatment, the co-inoculation of both pathogens led to similar symptoms as the inoculation of TC2-1 alone. However, for non-wound pre-treatment, which resemble the condition of natural infection, co-inoculation of both pathogens resulted in accelerated disease development and more severe symptoms compared with the inoculation of either pathogen individually (Figure 4).

To refine the taxonomic position of the pathogenic isolates, we performed MLSA analysis based on five housekeeping genes (atpD, avrBs2, cpn60, gyrB, and rpoD) for isolates TC1-1 and TC2-1, and four housekeeping genes (atpD, gyrB, infB, and rpoB) for isolates TC3-1 and TC3-2. As a result, the single-gene phylogenies of atpD, cpn60, gyrB, and rpoD, as well as the joint phylogenetic tree of all five genes, all contain a monophyletic clade consisting of TC1-1, TC2-1, and six other strains including X. axonopodis pv. citrumelo F1, X. axonopodis pv. commiphoreae strain LMG26789, X. perforans strains 91-118 and LH3, X. euvesicatoria LMG930, and X. campestris pv. vesicatirua 85-10 (Supplementary Figure 2). However, the phylogenetic results were inconclusive on the closest relatives of TC1-1 and TC2-1, leaving the taxonomic status of both strains at the species level uncertain. A high similarity between strains in different species indicates inconsistency in the current Xanthomonas classification situation. Therefore, we decided to sequence the whole genome of one of the isolates TC2-1 to determine the exact taxonomy of bacterial leaf canker pathogen of water spinach.

In addition, TC3-1 and TC3-2 have completely identical sequences of atpD, gyrB, infB, and rpoB, which are most similar to those of strain SGAir0210, sharing 99.11, 99.18, 99.38, and 99.36% identities, respectively. Joint phylogenetic analysis of atpD, gyrB, infB, and rpoB also indicated that strains TC3-1 and TC3-2 are grouped together with P. ananatis strains (Supplementary Figure 3). Thus, based on these results, we can conclude that strains TC3-1 and TC3-2 belong to Pantoea ananatis.

The TC2-1 genome was sequenced by a combination of Illumina short-read and Nanopore long-read sequencing technologies, and assembled into one circular chromosome (size: 5,093,283 bp; GC content: 64.76%) and two plasmids (sizes: 136,572 and 46,534 bp; GC contents: 60.96 and 62.10%) (Figure 5 and Table 1). A total of 3,848 protein-coding genes, 90 RNA genes (6 rRNAs, 51 tRNAs, and 30 ncRNAs), 671 pseudogenes, and 334 transposases (some are predicted as CDSs, some are pseudogenes) are annotated in the TC2-1 genome (GenBank no. CP077785.1) (Table 1).

The genome sequence of TC2-1 was analyzed with GTDB-Tk, which first placed TC2-1 in a pre-built bacterial phylogeny based on 120 proteins and then calculated the average nucleotide identity (ANI) values between TC2-1 and closely related type strains on the tree (Chaumeil et al., 2020). The result showed that TC2-1 is a member of X. perforans; the ANI value between TC2-1 and the type strain of X. perforans (strain CFBP 7293) is 97.78%, which is above the 95% threshold commonly used for bacterial species classification (Jain et al., 2018; Figure 6 and Supplementary Table 7). Additionally, the ANI values between TC2-1 and all 141 X. perforans genomes in the NCBI RefSeq database are above 97% (Supplementary Figure 4). Therefore, the water spinach pathogenic strain TC2-1 belongs to Xanthomonas perforans. To the best of our knowledge, water spinach is the fourth reported natural host plant of X. perforans besides tomato, pepper, and Eucalyptus pellita.

To identify TC2-1-specific genes related to host specialization, we performed a phylogenetic analysis of TC2-1 and all the 141 sequenced X. perforans strains in the NCBI RefSeq database based on 120 conserved, single-copy genes (Supplementary Figure 4) and selected five representative strains for further comparative analysis, including four strains isolated from tomato leaves (strain NI1 from Northern guinea savannah, and strains Xp11-2, GEV2392, and CFBP 7293 from different regions of the US) and one (strain AL66) isolated from pepper. Proteins present only in TC2-1, but not the other five strains, were then searched against the NCBI NR database to remove those with matches of high coverage (>80%) and identity (>65%) in any X. perforans strains. As a result, we identified a total of 277 genes that are unique to TC2-1, encoding 121 hypothetical proteins, 101 transposases, and 7 transcriptional regulators (Supplementary Table 8).

It is generally believed that there is a correlation between the repertoire of T3SEs and the host specificity of pathogenic Xanthomonas strains (Hajri et al., 2009, 2012a,b; Jacques et al., 2016). In the TC2-1 genome, we found an intact Type-III secretion system and predicted 56 T3SEs that are homologous to validated T3SEs in Xanthomonas and other bacteria (Table 2). Notably, 15 of these predicted T3SEs share low levels of sequence similarity with known T3SEs and, thus, were classified as TC2-1 unique. Among the predicted T3SEs, 19 encode short fragments of XOO4824 or XopT (Table 2), sharing high identity to each other; most of them were found to be adjacent to transposable elements, resulting in multiple copies in the genome. Actually, the real XopT and XOO4824 homologs in TC2-1 are TC2-1_004427 and TC2-1_004580, respectively, with 98.80% identity to each other. In summary, the number of predicted intact T3SEs in X. perforans strain TC2-1 is 37 indeed.